sensors

Article
Optimal Deployment of Vector Sensor Nodes in
Underwater Acoustic Sensor Networks
Sunhyo Kim 1 and Jee Woong Choi 2, *
1 Maritime Security Research Center, Korea Institute of Ocean Science and Technology, Busan 49111, Korea
2 Department of Marine Science and Convergence Engineering, Hanyang University-ERICA, Ansan,
Gyeonggi-do, Seoul 15588, Korea
* Correspondence: choijw@hanyang.ac.kr; Tel.: +82-31-400-5531




Received: 20 May 2019; Accepted: 27 June 2019; Published: 29 June 2019


Abstract: Underwater acoustic sensor networks have recently attracted considerable attention
as demands on the Internet of Underwater Things (IoUT) increase. In terms of efficiency, it is
important to achieve the maximum communication coverage using a limited number of sensor
nodes while maintaining communication connectivity. In 2017, Kim and Choi proposed a new
deployment algorithm using the communication performance surface, which is a geospatial
information map representing the underwater acoustic communication performance of a targeted
underwater area. In that work, each sensor node was a vertically separated hydrophone array, which
measures acoustic pressure (a scalar quantity). Although an array receiver is an effective system to
eliminate inter-symbol interference caused by multipath channel impulse responses in underwater
communication environments, a large-scale receiver system degrades the spatial efficiency. In this
paper, single-vector sensors measuring the particle velocity are used as underwater sensor nodes.
A single-vector sensor can be considered to be a single-input multiple-output communication system
because it measures the three directional components of particle velocity. Our simulation results
show that the optimal deployment obtained using single-vector sensor nodes is more effective than
that obtained using a hydrophone (three-channel vertical-pressure sensor) array.

Keywords: underwater acoustic sensor network; optimal deployment; vector sensor node; particle
velocity; communication performance surface

1. Introduction
Underwater acoustic sensor networks (UWASNs) are widely used in several applications such
as target detection and tracking, ocean pollution monitoring, and various ocean data collection
techniques [1–4]. The effectiveness of a UWASN can be evaluated in terms of the connectivity among
the sensor nodes deployed in a targeted area, as well as the coverage rate. It is hard to precisely
deploy numerous sensor nodes in underwater environments, and these nodes are also very expensive.
Therefore, in terms of efficiency, it is important to minimize the number of nodes deployed in a
targeted area.
Several approaches have been proposed to determine the optimal positions of sensor nodes in
UWASNs [5–7]. However, most attempts have assumed that the communication performance of
every node deployed at the target underwater area is the same without considering variations in the
ocean environment. Acoustic waves propagating through underwater channels undergo multiple
interactions with the ocean boundaries such as the sea surface and seabed, which cause severe time
delays, resulting in inter-symbol interference (ISI) [8]. ISI causes issues related to communication
demodulation. In addition, underwater communication is characterized by a time-varying channel,
which produces large Doppler spreading and a short coherence time [9–11]. In addition, marine

Sensors 2019, 19, 2885; doi:10.3390/s19132885 www.mdpi.com/journal/sensors

Sensors 2019, 19, 2885 2 of 10

organisms such as barnacles and algae might build up on a transducer as time goes on, which can
block the communication signals, causing data corruption. For these reasons, underwater acoustic
modems should be energy-efficient and resistant to biofouling [12].
Recently, Kim and Choi [13] suggested a new algorithm to maximize communication coverage
using a limited number of sensor nodes in shallow water areas. For this, we implemented a
communication performance surface (PS) in our algorithm. The communication PS is a geospatial map
displaying the communication performance information of the targeted area. The PS was obtained
based on channel impulse responses predicted by a BELLHOP model [14] for grid points in the targeted
area. The BELLHOP model is a highly efficient acoustic ray-tracing model for predicting acoustic
pressure fields in ocean environments.
Based on the PS, a virtual force-particle swarm optimization (VFPSO) algorithm was used to
determine the optimal positions of the sensor nodes. We assumed that each sensor node was a
vertical line array composed of vertically separated receiver elements to improve the communication
performance. The array processing techniques can be used to enhance the communication performance
by alleviating the effect of ISI [15,16]. However, such a large-scale array system degrades the spatial
efficiency in a UWASN.
In recent years, the propagation properties of acoustic vector quantities such as acceleration or
acoustic particle velocity have been studied [17–19]. Vector sensors have been applied in several
areas, including target detection and tracking, noise reduction, and acoustic communication [20–24].
A vector sensor measures the three directional components of the vector quantities, and accordingly, a
diversity gain can be achieved even when a single-vector sensor is used. In this paper, a vector sensor
measuring the particle velocity is used as a sensor node of a UWASN, with the aim of improving the
communication performance and the spatial efficiency. The outputs of the vector sensor are considered
to be the outputs of a single-input multiple-output (SIMO) communication system [22–24]. The optimal
algorithm proposed in our previous paper [13] is then applied to obtain the optimal deployment
scheme for the single-vector sensor nodes.
This paper is organized as follows. Section 2 provides a short summary of previous research and
describes the algorithm to calculate the communication PS of the single-vector sensor nodes. Section 3
describes the optimal deployment simulation for single-vector sensor nodes and compares the results
to that for the hydrophone (three-channel pressure sensor) array. Finally, Section 4 provides a summary
and conclusion.

2. Communication Performance Based on Single-Vector Sensors

2.1. Summary of Previous Research

Recently, Kim and Choi [13] suggested a methodology to maximize communication coverage
while maintaining connectivity among the sensor nodes for a given number of nodes. It was assumed
in most previous studies that the communication ranges of all underwater sensor nodes were identical.
In contrast, the algorithm proposed by Kim and Choi finds the best locations of the sensor nodes based
on a communication PS, which represents the spatial distribution of the relative performances in an
underwater communication system. The algorithm is composed of four subcategories: (1) underwater
acoustic channel modeling, (2) communication performance estimation, (3) the communication PS
construction, and (4) optimal nodal placement. In their study, the targeted area in the East Sea of
Korea was 22 km × 22 km and was divided into 100 grid points. The underwater acoustic channel
impulse responses, as a function of the distance between the source and receiver, were simulated
using the BELLHOP ray-tracing model with eight azimuthal angles for each grid point in the targeted
area. The acoustic parameters for the targeted underwater area (such as bathymetry, water sound
speed profile, and geoacoustic parameters for surficial sediments) are required to produce accurate
BELLHOP model outputs.
Sensors 2019, 19, x FOR PEER REVIEW 3 of 10
Sensors 2019, 19, 2885 3 of 10
We assumed a SIMO system consisting of sensor nodes in which each receiver was a 3‐channel
vertical line array. The communication signals received at every sensor array were simulated through
We assumed
convolution a SIMO
of the system
original consisting
binary of sensor
phase‐shift keying nodes
(BPSK)in which each receiver
communication was a 3-channel
sequence with the
vertical line array. The communication signals received at every sensor array
estimated channel impulse responses. Then, the signals were decoded using multichannel combining were simulated through
convolution
and adaptiveofdecision
the original binary
feedback phase-shift
equalization keying
(DFE), and(BPSK)
then thecommunication sequence
bit error rate (BER) with the
was estimated.
estimated channel impulse responses. Then, the signals were decoded using
The communication range was defined as the range corresponding to a BER of 2%, which was used multichannel combining
and
as aadaptive
tolerancedecision
criterionfeedback equalization
for the sensor network.(DFE),
Theand then the bit error
communication ratewere
ranges (BER) was estimated.
averaged for the
The
eightcommunication
azimuthal angles range wasthen
and defined as the range
interpolated for corresponding
all grid pointstotoa create
BER ofthe2%,communication
which was usedPS. as
aLastly,
tolerance criterion for the sensor network. The communication ranges were
based on the PS map, the VFPSO method [25] was applied to obtain the optimal deployment averaged for the eight
azimuthal
scheme ofangles and then interpolated
the underwater sensor nodes. forThe
all grid points
details to create
of the optimalthedeployment
communication PS. Lastly,
procedure are based
given
on the PS map,
in Reference [13]. the VFPSO method [25] was applied to obtain the optimal deployment scheme of the
underwater sensor nodes. The details of the optimal deployment procedure are given in Reference [13].
2.2. Underwater Acoustic Communication Using Single‐Vector Sensors
2.2. Underwater Acoustic Communication Using Single-Vector Sensors
An acoustic vector sensor measures the particle motion produced by acoustic waves, which is a
vectorAnquantity
acoustic described
vector sensor
using measures the particleparticle
the displacement, motionvelocity,
produced andby acceleration,
acoustic waves, whicha
whereas
is a vector quantity described using the displacement, particle velocity, and
hydrophone measures the acoustic pressure, which is a scalar quantity. Most previous studies in acceleration, whereas
aunderwater
hydrophone measures
acoustic the acousticused
communication pressure, which is
hydrophone a scalar
arrays quantity.
to receive the Most previous studies
communication signals
in
[15,16], but this approach has poor spatial efficiency (Figure 1a). In contrast, a vector communication
underwater acoustic communication used hydrophone arrays to receive the sensor has good
signals
spatial [15,16], butbecause
efficiency this approach has poor
it measures thespatial efficiency (Figure
three directional vector1a). In contrast,
quantities a vector
as well as thesensor has
pressure
good spatial efficiency because it measures the three directional vector quantities
change at a single point. Accordingly, it can be used as a SIMO system, as shown in Figure 1b [22– as well as the pressure
change
24]. at a single point. Accordingly, it can be used as a SIMO system, as shown in Figure 1b [22–24].

Figure 1. Communication system layouts. (a) Vertical-pressure sensor array, (b) single-vector sensor.
Figure 1. Communication system layouts. (a) Vertical‐pressure sensor array, (b) single‐vector
sensor.
In this paper, it is assumed for convenience that the vector sensor measures particle velocity
and pressure simultaneously in two-dimensional (2-D) coordinates (i.e., horizontal and vertical
In this paper, it is assumed for convenience that the vector sensor measures particle velocity and
components). Thus, the vector sensor can be considered a one-by-three SIMO system for underwater
pressure simultaneously in two‐dimensional (2‐D) coordinates (i.e., horizontal and vertical
acoustic communication. Figure 2 shows a block diagram of a vector sensor communication system
components). Thus, the vector sensor can be considered a one‐by‐three SIMO system for underwater
for decoding the communication data. To recover the baseband waveform, the three components of
acoustic communication. Figure 2 shows a block diagram of a vector sensor communication system
the communication signal are multiplied by e−iωt , where ω is the angular frequency. Then the rest
for decoding the communication data. To recover the baseband waveform, the three components of
of the process is the same as that in Reference [13], i.e., the baseband data are passed through a low
the communication signal are multiplied by 𝑒  , where ω is the angular frequency. Then the rest of
pass filter and an adaptive DFE with a recursive least-squares (RLS) algorithm [26]. Then the data
the process is the same as that in Reference [13], i.e., the baseband data are passed through a low pass
from multi-channels are combined to improve the signal-to-noise ratio (SNR) by removing residual
filter and an adaptive DFE with a recursive least‐squares (RLS) algorithm [26]. Then the data from
ISI [15,26,27]. The BER performance is finally evaluated. See [24] for a detailed description of vector
multi‐channels are combined to improve the signal‐to‐noise ratio (SNR) by removing residual ISI
sensor communication using particle velocity and pressure signals.
[15,26,27]. The BER performance is finally evaluated. See [24] for a detailed description of vector
sensor communication using particle velocity and pressure signals.
Sensors 2019,
Sensors 2019, 19,
19, 2885
x FOR PEER REVIEW 44 of
of 10
10
Sensors 2019, 19, x FOR PEER REVIEW 4 of 10

Figure 2. Block diagram of a single‐input multiple‐output (SIMO) system using a single‐vector sensor.
𝑃 is an
Figure acoustic
2. Block
Block pressure.
diagram
diagram of 𝑃 and 𝑃 are
ofaasingle‐input
single-input the horizontal
multiple‐output
multiple-output and vertical
(SIMO)
(SIMO) systemcomponents
system using of the pressure
usingaasingle‐vector
single-vectorsensor.
sensor.
𝑃
equivalent
P M is an acoustic velocities,𝑃Prespectively.
particlepressure. 𝑃z are the
r and P the horizontal
horizontal and vertical components of the pressure
respectively.
equivalent particle velocities, respectively.
Figure 3 shows the communication performance (BER) as a function of receiver position,
Figure
estimated 33shows
Figurefrom shows thethe
communication
communication
a three‐channel performance
performance
vertical‐pressure (BER)(BER)
sensor as a function
array as of receiver
anda afunction of position,
single‐vector sensorestimated
receiver position,
in a 2‐D
from a
estimatedthree-channel
underwater from vertical-pressure
a three‐channel
environment sensor
vertical‐pressure
with acoustically array
hard and and
sensor a single-vector
array andFor
soft bottoms. sensor
a single‐vectorin a 2-D
easier comparisons, underwater
sensor ainPekeris
a 2‐D
environment
underwater
waveguide with withaacoustically
environment
water with
depth hard
of 50andmsoft
acoustically and bottoms.
harda and
water For
soft easier
bottoms.
sound comparisons,
Forof
speed easier a Pekeris
1500comparisons,
m/s waveguide
was assumed.a PekerisIn
with a
waveguidewater depth
with a of
water50 m and
depth aofwater
50 m sound
and aspeed
water of 1500
sound m/s was
speed
addition, it was assumed that the sound source was at a depth of 25 m and transmitted a BPSKofassumed.
1500 m/s In addition,
was assumed. it was
In
assumed
addition, that
communication thesequence
it was sound
assumed source
thatwas
with athe atsound
1 kbpsa depth of 25
source
bit rate and m aand
was10attransmitted
kHza depth a 25
BPSK
carrieroffrequency. communication
m and transmitted
The sequence
a BPSK
BER performance
with a 1 kbps bit
communication
was obtained rate and awith
assequence
a function 10
ofkHz carrier
a 1 kbps
receiver bitfrequency.
rate and
position aThe BER
10 kHz
by varying performance
carrier
the frequency.
receiver wasfrom
depth obtained
The BER as
to a45function
5 m performance
m at 5‐
of
wasreceiver position
obtainedand
m intervals, as aby by varying
function
varyingofthe the receiver
receiver depth from
position byrange
source‐receiver 5 m
varying
fromto 45
the m at 5-m
75receiver
m to 500 intervals,
depth
m atfrom and m to 45 mInatthe
by varying
25‐m5intervals. 5‐
source-receiver
m intervals,
pressure andrange
sensor by fromthe75the
varying
array, m to 500spacing
m at 25-m
source‐receiver
element intervals.
range
was m, In
1.5from 75 the
m pressure
which 500 msensor
tocorrespondedat 25‐m array,10thetimes
intervals.
to element
In the
spacing
pressure was
wavelength. 1.5 m,
sensor
The which
array, thecorresponded
communication element to 10 was
spacing
signals receivedtimes them,
1.5
at each wavelength. The communication
which corresponded
hydrophone channel wereto 10 times
simulated signals
the
using
received
wavelength.at each hydrophone
The communication
the method described channel
in Reference signalswere simulated
[13]. received using the
at each hydrophone
The underwater method
acoustic channel described
channel in Reference
were simulated
response [13].
using
at each channel
The
the
pointunderwater
method
was predictedacoustic
described by channel
in Reference
the response
ray‐based The at
[13]. BELLHOP eachpropagation
underwaterchannel point
acoustic was predicted
channel
model andresponse byatthe
was convolved eachray-based
channel
with a
BELLHOP
point was propagation
representationpredicted bymodel
of the BPSK and was convolved
thesequence.
ray‐based BELLHOPwith a representation
propagation model and of thewas
BPSK sequence.with a
convolved
representation of the BPSK sequence.

Figure 3. Communication
Figure 3. Communication performance (BER estimates) as aa function function of
of receiver
receiver position.
position. Left
Left and
and
right
right plots
Figureplots are
arefor
forcoarse
coarsesandy
3. Communication sandy and
andsilty bottoms,
silty
performance respectively.
bottoms,
(BER as aUpper
respectively.
estimates) andand
Upper
function oflower plots
lower
receiver correspond
plots Lefttoand
correspond
position. the
to
performance
right plots areresults
the performance for a sandy
results
for coarse three-channel
for aand siltypressure
three‐channel sensor
pressure
bottoms, array and
sensor
respectively. a single-vector
array
Upper and
and sensor, respectively.
a single‐vector
lower sensor,
plots correspond to
respectively.
the performance results for a three‐channel pressure sensor array and a single‐vector sensor,
A finite difference approximation method [28] was used to simulate the particle velocities received
respectively.
at theAvector
finite sensor. The approximation
difference v(t) can[28]
particle velocitymethod be estimated
was usedby
to the time integral
simulate of the velocities
the particle pressure
A finite
received at thedifference approximation
vector sensor. method
The particle 𝑣 twas
velocity[28] canused to simulate
be estimated thetime
by the particle velocities
integral of the
received
pressure at the vector
gradient. Thesensor. The
pressure particlecan
gradient velocity 𝑣 t can be estimated using
be approximately by thetwo
timepressure
integralsignals
of the
pressure
received gradient. The pressure
at two pressure sensorsgradient can be
placed close approximately
together. estimated
The particle using
velocity two pressure
is given by [24]: signals
received at two pressure sensors placed close together. The particle velocity is given by [24]:
Sensors 2019, 19, 2885 5 of 10

gradient. The pressure gradient can be approximately estimated using two pressure signals received at
two pressure sensors placed close together. The particle velocity is given by [24]:

Zt
1 P2 (τ) − P1 (τ)
v(t) = dτ (1)
ρ0 d
0

where P1 and P2 are the acoustic pressure signals at the two adjacent sensors, d is the sensor spacing
between two sensors, ρ0 is the water density, and τ is the time variable. The acoustic pressure is
equivalent to the particle velocity in a particular direction multiplied by the acoustic impedance,
which is ρ0 c, where c is water sound speed. Then, the acoustic pressure PM at the center between
two adjacent sensors, and the horizontal and vertical components (Pr , Pz ) of the pressure equivalent
particle velocities are calculated by:

P1 (t) + P2 (t)
PM (t) = , Pr (t) = −ρ0 c vr (t), Pz (t) = −ρ0 c vz (t) (2)
2
where vr and vz are the horizontal and vertical components of the particle velocities, respectively.
For the simulation of particle velocities, two points, which were 3 cm apart both horizontally and
vertically, were selected, and their center was considered to be the receiver point. Then the pressure
channel impulse responses for these two adjacent points were predicted by the BELLHOP propagation
model, and finally, PM , Pr , and Pz were calculated by Equation (2). The distance between the two
adjacent points was equivalent to one fifth of the acoustic wavelength.Φ.
For the comparison of BER performances shown in Figure 3, two different kinds of bottom types
were considered. The hard bottom was a coarse sandy sediment with a mean grain size of 1.0 Φ and
the soft bottom was a silty sediment with a mean grain size of 6.0 Φ where Φ = − log2 (d/d0 ), d and
d0 are the grain diameter in millimeters and a reference length of 1 mm, respectively]. In general, an
acoustically hard bottom causes severe ISI due to strong reflection from the water-bottom interface,
whereas in case of a soft bottom, the number of dominant multipaths is limited due to a large bottom
loss. Overall, the BER performances obtained by the single-vector sensor outperformed those obtained
by the vertical-pressure sensor array in both the hard- and soft-bottom cases (Figure 3). In addition,
the performances in silty sediment were better than those in sandy sediment due to the effect of ISI.

2.3. Communication PS Based on a Single-Vector Sensor

A method of optimal deployment using a communication PS has been suggested in Reference [13].
A communication PS shows the spatial distribution of the communication performance [29]. Therefore,
if a communication PS is used in the process to search the optimal deployment positions, the rapid
computation of the search process becomes possible. In this paper, the same targeted area specified in
Reference [13] was used to evaluate the optimal deployment algorithm of the single-vector sensor nodes
and to compare it with the results obtained using the pressure sensor array. The size of the targeted
area was again 484 km2 , of which the western boundary was ~10 km away from the east coast of Korea.
A detailed description of the ocean environment in the targeted area, including bathymetry, water
sound speed profiles, and sediment geoacoustic properties, is provided in Reference [13]. The targeted
area was split up into 100 grid points. The channel impulse responses as a function of the range
for eight different azimuths (every 45◦ from 0◦ North) of each grid point were simulated using the
BELLHOP propagation model. The environmental, communication, and geometrical parameters used
to extract the communication PS are shown in Table 1.
Sensors 2019, 19,
Sensors 2019, 19, 2885
x FOR PEER REVIEW 66 of
of 10
10

Table
Latitude direction 1. The parameters used 22
distance forkm
the communication PS simulation.Root Raised Cosine filter
Pulse shaping

Communication
Ocean Environmental Parameters
Wind speed Value 10 m/s Equalizer Value DFE (RLS)
Adaptive
Parameters
Month
Azimuth angle interval February 45° SymbolBER
number
criterion 35002%
Longitude direction distance 22 km Symbol rate 1000 sps
Grid points
Latitude direction distance 22 km 100 Pulse shaping Root Raised Cosine filter
Wind speed Modeling Parameters10 m/s
Channel Equalizer Adaptive DFE (RLS)
Value
Azimuth angle interval 45 ◦ BER criterion 2%
Frequency 10 kHz
Grid points 100
Source level
Channel Modeling Parameters Value140 dB
Frequency
Source depth 2 10 kHz the bottom
m above
Source level Sensor depth 140mdB
0.5–3.5 above the bottom
Source depth 2 m above the bottom
Three‐channel vertical‐pressure
Three-channel sensor arraySensor depth
vertical-pressure 0.5–3.5 m above the(10
bottom
Element spacing 1.5 m λ)
sensor array Element spacing 1.5 m (10 λ)
Single-vector
Single‐vectorsensor
sensor Sensor depth
Sensor depth 2 m above the bottom
2 m above the bottom

simulated channel
The simulated channel impulse
impulse responses
responses were
were convolved
convolved withwith the
the original
original communication
communication
sequence to to simulate
simulatethe thecommunication
communication signals. Then,
signals. BERBER
Then, estimates werewere
estimates performed for thefor
performed three‐
the
channel vertical‐pressure sensor array and the single‐vector sensor using the demodulation
three-channel vertical-pressure sensor array and the single-vector sensor using the demodulation process
shown, shown,
process as shown as in Figure
shown in4Figure
in Reference [13] and[13]
4 in Reference Figure
and 2Figure
in this2paper,
in this respectively. Figure Figure
paper, respectively. 4 shows 4
the BER
shows theperformances
BER performances interpolated for the
interpolated for eight azimuths
the eight of each
azimuths of eachgrid
gridpoint
pointfor
forthe
thetwo
two receiver
systems. The
The results
results show
show that the single-vector
single‐vector sensor performed better than the the three-channel
three‐channel
vertical‐pressure sensor array.
vertical-pressure array. Based on these results, the range corresponding to a BER of 2% was was
defined as the communication range, as in our previous
defined as the communication range, as in our previous work [13]. work [13].

Figure performance field for 100 grid points predicted for (a) the three‐channel
Figure 4. BER performance three-channel pressure sensor
array, and (b) the single-vector sensor in the targeted area in February.
array, and (b) the single‐vector sensor in the targeted area in February.

Then
Then the
the communication
communication ranges
ranges forfor the
the eight
eight azimuths
azimuths were
were averaged,
averaged, and
and the
the mean
mean value
value was
was
defined
defined as the communication radius for each grid point, assuming that the communication radius
as the communication radius for each grid point, assuming that the communication radius
represents
represents the
the communication
communication performance
performance for for the
the grid
grid point. Finally, the
point. Finally, the communication
communication PS PS was
was
obtained
obtained by an interpolation of the communication radii for every grid point. Figure 5a,b show the
by an interpolation of the communication radii for every grid point. Figure 5a,b show the
communication
communication PS PS simulated
simulated for
for the
the three-channel
three‐channel pressure
pressure sensor
sensor array
array and
and for
for the
the single-vector
single‐vector
sensor,
sensor, respectively.
respectively. The
The results
results also
also imply
imply that
that the
the single-vector
single‐vector sensor
sensor has
has aa better
better communication
communication
range compared with the three-channel pressure sensor
range compared with the three‐channel pressure sensor array. array.
Sensors 2019, 19, 2885 7 of 10
Sensors 2019, 19, x FOR PEER REVIEW 7 of 10

Figure
Figure 5. Communication PS
5. Communication PS simulated
simulated for
for the
the targeted
targeted area
area simulated
simulated for
for (a)
(a) the
the three-channel
three‐channel
pressure sensor array, and (b) the single-vector sensor in the targeted area in February.
pressure sensor array, and (b) the single‐vector sensor in the targeted area in February.
3. Optimal Deployment of Underwater Sensor Nodes
3. Optimal Deployment of Underwater Sensor Nodes
The optimization algorithm we used for sensor node deployment was the VFPSO [25], as used
The optimization algorithm we used for sensor node deployment was the VFPSO [25], as used
in Reference [13]. This algorithm is a hybrid method that combines the advantages of the virtual
in Reference [13]. This algorithm is a hybrid method that combines the advantages of the virtual force
force algorithm (VFA) [30,31] and particle swarm optimization (PSO) [32,33], which was also used
algorithm (VFA) [30,31] and particle swarm optimization (PSO) [32,33], which was also used in
in Reference [13]. The PSO can find a global optimal position by comparing the particle position
Reference [13]. The PSO can find a global optimal position by comparing the particle position of the
of the current generation to the experiences achieved at previous generations, and the VFA has an
current generation to the experiences achieved at previous generations, and the VFA has an
outstanding ability to adjust the distance between sensor nodes using attractive and repulsive forces.
outstanding ability to adjust the distance between sensor nodes using attractive and repulsive forces.
Therefore, in this case the VFPSO was suitable to search the optimal positions of sensor nodes where the
Therefore, in this case the VFPSO was suitable to search the optimal positions of sensor nodes where
communication coverage rate was maximized while ensuring that the connectivity of the sensor nodes
the communication coverage rate was maximized while ensuring that the connectivity of the sensor
was maintained. A detailed description of the VFPSO algorithm for optimal sensor node deployment
nodes was maintained. A detailed description of the VFPSO algorithm for optimal sensor node
is provided in Reference [13].
deployment is provided in Reference [13].
In this paper, one hundred sensor nodes were first randomly placed in the targeted area, and
In this paper, one hundred sensor nodes were first randomly placed in the targeted area, and
the VFPSO algorithm was then applied to the three-channel vertical-pressure sensor array and the
the VFPSO algorithm was then applied to the three‐channel vertical‐pressure sensor array and the
single-vector sensor, based on the PS obtained in Section 2. The parameters used in the VFPSO
single‐vector sensor, based on the PS obtained in Section 2. The parameters used in the VFPSO
algorithm for the optimal deployment simulation are given in Table 2.
algorithm for the optimal deployment simulation are given in Table 2.
Table 2. The parameters used in the simulation for optimal sensor node deployment.
Table 2. The parameters used in the simulation for optimal sensor node deployment.
Optimal Deployment Parameters Value
Optimal Deployment
Loop number Value
50
Parameters
Sensor node number 100
Loopforce
Attractive number
weight 50
0.01
Sensor node
Repulsive force number
weight 100
0.5
Acceleration
Attractive weight
force weight 1
0.01
Repulsive force weight 0.5
Figure 6 shows one hundred sensor Acceleration weight optimally1 in the targeted area, and their
nodes deployed
connectivity, which were displayed on their communication PS. The efficiency of the optimal deployment
couldFigure
then be6 shows one hundred
investigated using thesensor nodes deployed
communication optimally
coverage in the was
rate, which targeted area,
defined asand their
the ratio
connectivity,
of the area of which were displayed
communication coverage onbytheir communication
the sensor PS. of
nodes to that Thetheefficiency of the
targeted area optimal
[13]. As a
deployment could then be investigated using the communication coverage rate, which
result, the communication coverage rate simulated using the single-vector sensor nodes was estimated was defined
as be
to the95.3%,
ratio ofwhereas
the areathat
of communication coverage
of the three-channel by thesensor
pressure sensornodes
nodeswas to that of the
85.2%. targeted
This result area
was
[13]. As a result, the communication coverage rate simulated using the single‐vector
expected because the single-vector sensor had better communication performance compared to the sensor nodes
was estimatedpressure
three-channel to be 95.3%, whereas
sensor that
array, as of the three‐channel
mentioned pressure sensor nodes was 85.2%. This
in Section 2.3.
result was expected because the single‐vector sensor had better communication performance
compared to the three‐channel pressure sensor array, as mentioned in Section 2.3.
Sensors 2019, 19, 2885 8 of 10
Sensors 2019, 19, x FOR PEER REVIEW 8 of 10

Figure 6. Results of the optimal deployment simulation using two different sensor node types: (a) the
Figure 6. Results of the optimal deployment simulation using two different sensor node types: (a) the
three-channel vertical-pressure sensor array, and (b) the single-vector sensor. The solid lines between
three‐channel vertical‐pressure sensor array, and (b) the single‐vector sensor. The solid lines between
the sensor nodes indicate their connectivity, which was estimated by comparing the communication
the sensor nodes indicate their connectivity, which was estimated by comparing the communication
range of each sensor node with the range from each sensor node to its neighboring nodes. In both cases,
range of each sensor node with the range from each sensor node to its neighboring nodes. In both
every sensor node was 100% connected.
cases, every sensor node was 100% connected.
4. Summary and Conclusions
4. Summary and Conclusions
Most previous studies on underwater acoustic communication used a multichannel underwater
Most previous
communication studies
system on underwater
with an array ofacoustic
spatially communication used a multichannel
separated hydrophone receivers underwater
to increase
communication system with an array of spatially separated hydrophone
the communication performance. Our previous paper [13] proposed a methodology of optimal receivers to increase the
communication performance. Our previous paper [13] proposed
deployment of underwater sensor nodes, which employed hydrophone arrays. Our algorithm used a methodology of optimal
deployment
the communicationof underwater
PS, which sensor nodes,the
can reflect which
effectsemployed
of ocean hydrophone
environmental arrays. Our algorithm
variations on underwater used
the communication PS, which can reflect the effects of ocean environmental
communication. However, a large-sized receiver system compromises the spatial efficiency of a variations on underwater
communication.
UWASN system. However, a large‐sized receiver system compromises the spatial efficiency of a
UWASN This system.
paper considers a single-vector sensor that measures particle velocities as an underwater
sensorThis
node paper
in a considers
UWASN system, a single‐vector
instead of sensor that measures
a hydrophone array particle
measuring velocities
acoustic aspressure,
an underwater
which
sensor node in a UWASN system, instead of a hydrophone array measuring
is a scalar quantity. Since a vector sensor measures the three directional components of acoustic acoustic pressure, which
is a scalar
vector quantity.
quantities, it Since
can bea vector
used assensor
a SIMO measures
system.the three directional
Therefore, diversitycomponents
gain can be of acousticduring
achieved vector
quantities, it can be used as a SIMO system. Therefore, diversity gain
the decoding process of underwater acoustic communication, even with a single-vector sensor. Our can be achieved during the
decodingtoprocess
method determine of underwater
the optimal acoustic
deployment communication,
of sensor nodes even(except
with afor single‐vector
the difference sensor. Our
of signal
method to determine the optimal deployment of sensor nodes (except for
demodulation process owing to sensor type) was the same as in our previous work [13]. As a result, in the difference of signal
demodulation
our simulation,process owing to sensor
the performance type) was the
of a single-vector samewas
sensor as inbetter
our previous
than thatworkof the [13]. As a result,
three-channel
in our simulation,sensor
vertical-pressure the performance of a single‐vector
array. Specifically, sensor was better
the communication thanrate
coverage that obtained
of the three‐channel
using one
hundred single-vector sensors (which was 95.3%) was a ~10% improvement over obtained
vertical‐pressure sensor array. Specifically, the communication coverage rate that obtained using one
using
hundred single‐vector sensors (which was 95.3%) was a ~10% improvement
one hundred three-channel vertical-pressure sensor arrays (which was 85.2%). The simulation was over that obtained using
one hundred
further carriedthree‐channel
out and the resultsvertical‐pressure
indicated that sensor arrays (which was
the communication 85.2%).rate
coverage Theforsimulation
one hundred was
further carried out and the results indicated that the communication
three-channel hydrophone arrays could be achieved using 80 single-vector sensor nodes, with a coverage rate for one hundred
three‐channel
coverage rate of hydrophone arrays could
84.7%. In addition, when bethe achieved
simulationusingwas80 repeated
single‐vector
usingsensor nodes, with
one hundred singlea
coverage rate of 84.7%. In addition, when the simulation
pressure sensor nodes, the coverage rate was estimated to be only 45.6%. was repeated using one hundred single
pressure sensor nodes, the coverage rate was estimated to be only 45.6%.
In this paper, we proposed an algorithm for determining the optimal locations of underwater
sensorInnodes
this paper,
underwe theproposed
assumption an algorithm for determining
that each sensor node was the optimalof
composed locations of underwater
a single-vector sensor.
sensor nodes under the assumption that each sensor node was composed
If a vector sensor array is used as a sensor node, the spatial efficiency might be further improved, of a single‐vector sensor. If
a vector
thus sensor better
achieving array is used as a sensor
communication node, the spatial
performance efficiency
and larger might be further
communication improved,
coverage. thus
However,
achieving
both better communication
the advantages and disadvantagesperformanceshouldand larger communication
be considered with respectcoverage.
to spatialHowever,
efficiency both
and
the advantages
cost efficiency. and disadvantages should be considered with respect to spatial efficiency and cost
efficiency.
Author Contributions: J.W.C. designed and led the research, and S.K. applied the algorithm to the UWASN and
performed the result analysis.
Author Contributions: J.W.C. Both authors
designed andwrote the
led the paper. The
research, andauthors declare
S.K. applied theno conflictsto
algorithm ofthe
interest.
UWASN and
performed the result analysis. Both authors wrote the paper. The authors declare no conflicts of interest.
Sensors 2019, 19, 2885 9 of 10

Funding: This work was supported by Agency for Defense Development, Korea, under grant UD170006DD,
the National Research Foundation of Korea (NRF-2016R1D1A1B03930983), a “Development of the wide-area
underwater mobile communication systems” grant (PMS4110) from National R&D Project, funded by Ministry of
Oceans and Fisheries (MOF) and a “Development of maritime defense and security technology” grant (PE99741),
promoted by the Korea Institute of Ocean Science and Technology (KIOST).
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Akyildiz, I.F.; Pompili, D.; Melodia, T. Underwater acoustic sensor networks: Research challenges.
Ad. Hoc. Netw. 2005, 3, 257–279. [CrossRef]
2. Lloret, J. Underwater sensor nodes and networks. Sensors 2013, 13, 11782–11796. [CrossRef] [PubMed]
3. Pompili, D.; Melodia, T.; Akyildiz, I.F. Three-dimensional and two-dimensional deployment analysis for
underwater acoustic sensor networks. Ad Hoc Netw. 2009, 7, 778–790. [CrossRef]
4. Akkaya, K.; Newell, A. Self-deployment of sensors for maximized coverage in underwater acoustic sensor
networks. Comput. Commun. 2009, 32, 1233–1244. [CrossRef]
5. Kilfoyle, D.B.; Baggeroer, A.B. The State of the Art in Underwater Acoustic Telemetry. IEEE J. Ocean. Eng.
2000, 25, 4–27. [CrossRef]
6. Du, H.; Xia, N.; Zheng, R. Particle swarm inspired underwater sensor self-deployment. Sensors 2014, 14,
15262–15281. [CrossRef] [PubMed]
7. Luo, H.; Guo, Z.; Dong, W.; Hong, F.; Zhao, Y. LDB. Localization with directional beacons for sparse 3D
underwater acoustic sensor networks. J. Netw. 2010, 5, 28–38. [CrossRef]
8. Choi, J.W.; Dahl, P.H. Measurement and simulation of the channel intensity impulse response for a site in the
East China Sea. J. Acoust. Soc. Am. 2006, 119, 2677–2685. [CrossRef]
9. Rouseff, D.; Badiey, M.; Song, A. Effect of reflected and refracted signals on coherent underwater acoustic
communication: Results from the Kauai experiment (KauaiEx 2003). J. Acoust. Soc. Am. 2009, 126, 1359–2366.
[CrossRef]
10. Song, A.; Badey, M.; Song, S.C.; Hodgkiss, S.; Porter, M.B. Impact of ocean variability on coherent underwater
acoustic communications during the Kauai experiment (KauaiEx). J. Acoust. Soc. Am. 2008, 123, 856–865.
[CrossRef]
11. Kim, S.; Son, S.U.; Kim, H.; Choi, K.H.; Choi, J.W. Estimate of Passive Time Reversal Communication
Performance in Shallow Water. Appl. Sci. 2018, 8, 23. [CrossRef]
12. Sendra, S.; Lloret, J.; Jimenez, J.M.; Parra, L. Underwater acoustic modems. IEEE J. Sens. 2016, 16, 4063–4071.
[CrossRef]
13. Kim, S.; Choi, J.W. Optimal Deployment of Sensor Nodes Based on Performance Surface of Underwater
Acoustic Communication. Sensors 2017, 17, 2389–2403.
14. Porter, M.B.; Bucker, H.P. Gaussian beam tracing for computing ocean acoustic fields. J. Acoust. Soc. Am.
1987, 82, 1349–1359. [CrossRef]
15. Stojanovic, M.; Catipovic, J.; Proakis, J.G. Adaptive multichannel combining and equalization for underwater
acoustic communications. J. Acoust. Soc. Am. 1993, 94, 1621–1631. [CrossRef]
16. Song, H.C.; Hodgkiss, S.; Kuperman, W.A.; Higley, W.J.; Raghukumar, K.; Akal, T. Spatial diversity in passive
time reversal communications. J. Acoust. Soc. Am. 2006, 120, 2067–2076. [CrossRef]
17. Dall’Osto, D.R.; Dahl, P.H.; Choi, J.W. Properties of the acoustic intensity vector field in a shallow water
waveguide. J. Acoust. Soc. Am. 2012, 131, 2023–2035. [CrossRef] [PubMed]
18. D’Spain, G.L.; Luby, J.C.; Wilson, G.R.; Gramann, R.A. Vector sensors and vector sensor line arrays: Comments
on optimal array gain and detection. J. Acoust. Soc. Am. 2006, 120, 171–185. [CrossRef]
19. Dall’Osto, D.R.; Choi, J.W.; Dahl, P.H. Measurement of acoustic particle motion in shallow water and its
application to geoacoustic inversion. J. Acoust. Soc Am. 2016, 139, 311–319. [CrossRef]
20. Hawkes, M.; Nehorai, A. Acoustic Vector-Sensor Correlations in Ambient Noise. IEEE J. Ocean. Eng. 2001,
26, 337–347. [CrossRef]
21. Cray, B.A.; Nuttall, A.H. Directivity factors for linear arrays of velocity sensors. J. Acoust. Soc. Am. 2001, 110,
324–331. [CrossRef]
22. Song, A.; Abdi, A.; Badiey, M.; Hursky, P. Experimental Demonstration of Underwater Acoustic
Communication by Vector Sensors. IEEE J. Ocean. Eng. 2011, 36, 454–461. [CrossRef]
Sensors 2019, 19, 2885 10 of 10

23. Abdi, A.; Guo, H. A New Compact Multichannel Receiver for Underwater Wireless Communication
Networks. IEEE Trans. Commun. 2009, 8, 3326–3329. [CrossRef]
24. Kim, S.; Kim, H.; Jung, S.; Choi, J.W. Time reversal communication using vertical particle velocity and
pressure signal in shallow water. Ad. Hoc. Netw. 2019, 89, 161–169. [CrossRef]
25. Wang, X.; Wang, S.; Ma, J.J. An improved co-evolutionary particle swarm optimization for wireless sensor
networks with dynamic deployment. Sensors 2007, 7, 354–370. [CrossRef]
26. Proakis, J.G. Digital Communications; McGraw Hill: New York, NY, USA, 2008; pp. 298–315.
27. Stojanovic, M.; Catipovic, J.A.; Proakis, J.G. Phase-Coherent Digital Communications for Underwater
Acoustic Channels. IEEE J. Ocean. Eng. 1993, 19, 100–111. [CrossRef]
28. Fahy, F.J. Sound Intensity; E&FN Spon: London, UK, 1995; pp. 1053–1075.
29. McDowell, P. Environmental and Statistical Performance Mapping Model for Underwater Acoustic Detection
Systems. Ph.D. Thesis, University of New Orleans, New Orleans, LA, USA, 2010.
30. Chen, J.; Li, S.; Sun, Y. Novel Deployment Schemes for Mobile Sensor Networks. Sensors 2007, 7, 2907–2919.
[CrossRef]
31. Zou, Y.; Chakrabarty, K. Sensor deployment and target localization based on virtual forces. In Proceedings
of the IEEE INFOCOM 2003. Twenty-Second Annual Joint Conference of the IEEE Computer and
Communications Societies (IEEE Cat. No.03CH37428), San Francisco, CA, USA, 30 March–3 April 2003.
32. Majid, A.S.; Joelianto, E. Optimal Sensor Deployment in Non-Convex Region using Discrete Particle Swarm
Optimization Algorithm. In Proceedings of the IEEE Conference on Control, Systems & Industrial Informatics,
Bandung, Indonesia, 23–26 September 2012.
33. Wang, X.; Ma, J.; Wang, S.; Bi, D.W. Distributed particle swarm optimization and simulated annealing for
energy-efficient coverage in wireless sensor networks. Sensors 2007, 7, 628–648. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).